(1) QC — glidertest diagnostics
- Flat-line test: flags a stretch where the value barely changes — its range (max − min) over a trailing window (the suspect/fail window, in seconds) stays within the tolerance. A longer window or a tighter tolerance flags fewer points, so a stuck sensor is caught but homogeneous water is not.
- Spike test: flags a lone point that jumps away from its neighbours — the value minus the average of its two adjacent samples exceeds the threshold, catching a single outlier rather than a sustained offset.
Profile number: No issues detected. Profile duration: 55 profiles have abnormal duration.
glidertest diagnostics — run on the fly, not written to the file. Thresholds are glidertest's defaults unless a report config overrides them, shown so the verdict can be reproduced.
| Variable | Test | Suspect range / threshold | Fail range / threshold | Source |
|---|---|---|---|---|
| TEMP | gross-range | [0, 30] | [-2.5, 40] | default |
| TEMP | spike | |Δ| > 2.0 | |Δ| > 6.0 | default |
| TEMP | flat-line | flat ≥ 30 s (tol ±0.0005) | flat ≥ 60 s (tol ±0.0005) | default |
| TEMP | hysteresis | > 5% dive–climb error per depth bin | > 5 bins exceed | default |
| PSAL | gross-range | [5, 38] | [2, 41] | default |
| PSAL | spike | |Δ| > 0.3 | |Δ| > 0.9 | default |
| PSAL | flat-line | flat ≥ 30 s (tol ±0.0005) | flat ≥ 60 s (tol ±0.0005) | default |
| PSAL | hysteresis | > 5% dive–climb error per depth bin | > 5 bins exceed | default |
| DOXY | gross-range | [0, 350] | [0, 500] | default |
| DOXY | spike | |Δ| > 10 | |Δ| > 50 | default |
| DOXY | flat-line | flat ≥ 30 s (tol ±0.01) | flat ≥ 60 s (tol ±0.01) | default |
| DOXY | hysteresis | > 5% dive–climb error per depth bin | > 5 bins exceed | default |
| CHLA | gross-range | [0, 15] | [-1, 20] | default |
| CHLA | spike | |Δ| > 1 | |Δ| > 5 | default |
| CHLA | flat-line | flat ≥ 30 s (tol ±0.01) | flat ≥ 60 s (tol ±0.01) | default |
| CHLA | hysteresis | > 5% dive–climb error per depth bin | > 5 bins exceed | default |
Test × variable: flagged-sample counts (per-sample tests show a small dashed bar). Gross range applies the suspect span only — the fail span above is not yet evaluated. Hysteresis counts depth bins with >5% dive–climb error; a variable is flagged when more than 5 bins exceed it.
| Test | TEMP | PSAL | DOXY | CHLA |
|---|---|---|---|---|
| Gross range | clean | 241 susp 1,365 fail | 256,448 susp 1,352,678 fail | 681 susp 638 fail |
| Spike | clean | 450 susp 876 fail | 194 susp | 4,247 susp 340 fail |
| Flat line | 6,973 susp 1,611 fail | 72 susp 21 fail | clean | 1,660 susp 1,490 fail |
| Hysteresis (mean) | 0/176 bins | 0/176 bins | 13/176 bins | 2/176 bins |
| Hysteresis (range) | 0/176 bins | 0/176 bins | 7/176 bins | 0/176 bins |
(2) Gross range (T–S)
Current gross-range limits (applied)
- TEMP — suspect [0, 30], fail [-2.5, 40]
- PSAL — suspect [5, 38], fail [2, 41]
Suggested gross-range limits (this deployment) — copy into your config
Each span is the central 99.9% of the finite values padded by 1% (suspect) and 5% (fail) of that span. Users should verify that the recommended choices are appropriate for their dataset.
(3) QC — TEMP
Current limits (applied)
- TEMP flat line — suspect ≥ 30 s, fail ≥ 60 s (tol ±0.0005)
- TEMP spike — suspect Δ > 2, fail Δ > 6
Suggested limits (this deployment) — copy into your config
TEMP is sampled about every 1 s; a flat-line window needs to span a couple of its samples, so the suggested suspect is 10× and fail 20× that interval (a window shorter than the sampling flags every point). This is a starting point: if it flags too many points as flat-line, lengthen the window (raise suspect_threshold / fail_threshold) or lower the tolerance. The TEMP spike thresholds are not mission-derived; the applied values are shown so you can loosen them (raise suspect_threshold / fail_threshold) if real features are flagged. Users should verify that the recommended choices are appropriate for their dataset.
Flat-line runs
Spike
(4) QC — PSAL
Current limits (applied)
- PSAL flat line — suspect ≥ 30 s, fail ≥ 60 s (tol ±0.0005)
- PSAL spike — suspect Δ > 0.3, fail Δ > 0.9
Suggested limits (this deployment) — copy into your config
PSAL is sampled about every 1 s; a flat-line window needs to span a couple of its samples, so the suggested suspect is 10× and fail 20× that interval (a window shorter than the sampling flags every point). This is a starting point: if it flags too many points as flat-line, lengthen the window (raise suspect_threshold / fail_threshold) or lower the tolerance. The PSAL spike thresholds are not mission-derived; the applied values are shown so you can loosen them (raise suspect_threshold / fail_threshold) if real features are flagged. Users should verify that the recommended choices are appropriate for their dataset.
Flat-line runs
Spike
(5) QC — DOXY
Current limits (applied)
- DOXY gross range — suspect [0, 350], fail [0, 500]
- DOXY flat line — suspect ≥ 30 s, fail ≥ 60 s (tol ±0.01)
- DOXY spike — suspect Δ > 10, fail Δ > 50
Suggested limits (this deployment) — copy into your config
The gross span is the central 99.9% of the finite values ([-4.1, 389.21]) padded by 1% (suspect) and 5% (fail) of that span (lower bound clamped at 0). DOXY is sampled about every 1 s; a flat-line window needs to span a couple of its samples, so the suggested suspect is 10× and fail 20× that interval (a window shorter than the sampling flags every point). This is a starting point: if it flags too many points as flat-line, lengthen the window (raise suspect_threshold / fail_threshold) or lower the tolerance. The DOXY spike thresholds are not mission-derived; the applied values are shown so you can loosen them (raise suspect_threshold / fail_threshold) if real features are flagged. Users should verify that the recommended choices are appropriate for their dataset.
Flat-line runs
DOXY: flat-run durations against the test windows and the median sampling interval — applicable but unavailable: ValueError: DOXY: no flat-line flag to plot
DOXY: start-to-end change of each flat run against the tolerance (stuck sensor vs slow signal) — applicable but unavailable: ValueError: DOXY: no multi-sample flat run (flat flags are all isolated samples)
DOXY: zoom on the longest flagged flat run, with the ±tolerance band — applicable but unavailable: ValueError: DOXY: no flat-line flag to plot
Spike
(6) QC — CHLA
Current limits (applied)
- CHLA gross range — suspect [0, 15], fail [-1, 20]
- CHLA flat line — suspect ≥ 30 s, fail ≥ 60 s (tol ±0.01)
- CHLA spike — suspect Δ > 1, fail Δ > 5
Suggested limits (this deployment) — copy into your config
The gross span is the central 99.9% of the finite values ([0.05, 3.78]) padded by 1% (suspect) and 5% (fail) of that span (lower bound clamped at 0). CHLA is sampled about every 1 s; a flat-line window needs to span a couple of its samples, so the suggested suspect is 10× and fail 20× that interval (a window shorter than the sampling flags every point). This is a starting point: if it flags too many points as flat-line, lengthen the window (raise suspect_threshold / fail_threshold) or lower the tolerance. The CHLA spike thresholds are not mission-derived; the applied values are shown so you can loosen them (raise suspect_threshold / fail_threshold) if real features are flagged. Users should verify that the recommended choices are appropriate for their dataset.